Attackers Abuse Microsoft Teams and AD CS to Steal Data

Dominic Jainy stands at the forefront of modern identity security, bringing years of deep-tier expertise in navigating the complex intersection of human psychology and enterprise infrastructure. As an authority on threat intelligence, he has spent his career dissecting how high-level adversaries bypass the most expensive firewalls by exploiting the simplest of human trust. Today, he joins us to discuss a recent, alarming trend in which attackers are moving beyond simple phishing to orchestrate sophisticated “hands-on” intrusions. By blending social engineering on platforms like Microsoft Teams with the technical subversion of Active Directory Certificate Services, these threat actors have found a way to turn a routine help-desk call into a total domain compromise.

The following discussion explores the lifecycle of these modern identity-based attacks, focusing on the mechanics of initial access and the exploitation of long-standing vulnerabilities in certificate management. We examine how attackers leverage remote-support tools to drop malicious payloads, the specific weaknesses known as ESC1 and ESC8 that allow for privileged escalation, and the critical lessons learned when automated containment systems leave even a single door unlocked. This is a story of how old techniques, some dating back several years, remain lethal in the hands of persistent adversaries in 2026.

Help desk impersonation often involves using external communication platforms to trick employees into sharing remote access. How do these attackers successfully masquerade as legitimate support, and what makes the use of Quick Assist particularly effective for them?

The brilliance of this attack lies in its sheer simplicity and the exploitation of established corporate trust. Attackers create external Microsoft Teams tenants with names that sound like an official “Support” or “IT Help Desk” department, making the incoming notification look entirely routine to a busy employee. When that message pops up, or the phone rings through the Teams interface, the victim isn’t thinking about a nation-state actor; they’re thinking about fixing a technical glitch so they can get back to their work. The attackers lean into this helpful persona, guiding the user to open Quick Assist, which is a legitimate tool baked right into Windows. Because it’s a trusted, native application, it doesn’t trigger the same “stranger danger” alarms that a third-party download might. Once the employee reads that connection code aloud, the attacker isn’t just looking at the screen; they have full, hands-on control of the workstation. It’s a visceral moment—the employee watches their mouse move as if by a ghost, believing a colleague is helping them, while in reality, the attacker is already preparing to drop their toolkit into the Downloads folder.

Once an intruder has a foothold via a remote-help tool, they often deploy custom runtimes to execute their scripts. What did we see in these specific intrusions regarding the payloads, and why is it so difficult for traditional file-name monitoring to catch them?

The attackers are incredibly agile when it comes to their execution environment, often bypassing the limitations of built-in scripting by bringing their own runtimes. In the cases we’ve analyzed, they dropped archives that looked like innocuous Windows updates, but inside were portable versions of Python, Bun, or Java. By using these portable runtimes, they can run sophisticated attack scripts—like the Python scripts we’ve identified with hashes beginning in f326 and 06c5—without needing to install anything that would require admin privileges or trigger a standard “new software” alert. They are very clever with naming; they’ll rename a file to look like a system component, even while the metadata still reveals the original runtime version. This makes file-name monitoring almost useless because the “Update.exe” you see running might actually be a renamed Java environment executing a credential-stealing script. It’s a game of shadows where the attacker is using the system’s own flexibility against it, making the malicious activity blend in with the background noise of a busy OS.

The move from local workstation control to Active Directory Certificate Services (AD CS) is a sophisticated jump. Could you break down how a single misconfigured certificate template, like the ESC1 weakness, essentially hands over the keys to the entire kingdom?

Active Directory Certificate Services is the heart of identity for many organizations, and that makes it a prime target. The ESC1 vulnerability is particularly terrifying because it stems from a template configuration that allows a requester to supply a “Subject Alternative Name” or SAN. In plain English, this means a low-privileged user can ask the server for a login certificate and say, “By the way, I’m actually the Domain Administrator.” If the template is poorly configured, the server believes them and issues a certificate that proves that false identity. Once the attacker has that certificate, they don’t need a password, and they don’t care if you change the password later. They can use that certificate to log in as a high-privileged account from anywhere. In the first case we saw, the attackers hit a small group of users and instantly generated certificates for three highly privileged accounts. It’s a total bypass of the standard identity hierarchy, turning a regular employee’s workstation into a mint that prints “get out of jail free” cards for the entire network.

In one of the documented cases, an attacker managed to exfiltrate data despite automatic defenses kicking in. Why did that specific containment strategy fail, and what does it teach us about the dangers of “partial identity containment”?

This is one of the most critical takeaways for incident responders. In the first intrusion, Microsoft Defender’s automatic disruption worked remarkably well at first—it identified the suspicious activity and cut off the sessions for most of the compromised identities. However, one single privileged account was manually or accidentally excluded from that automatic response. That one account was the “opening” the attacker needed. While the system was busy blocking 1,000 other attempts to access files or log in via SMB, the attacker sat quietly with that one overlooked account. About 90 minutes after the first wave, they launched a second, larger wave targeting 16 users and managed to grab 8 more privileged certificates. Because they had that one viable route, they could reach the file server, run a renamed version of Rclone, and start funnelling company data to an outside SFTP server. It teaches us that in identity security, 99% containment is 0% security. If you leave even one path open, a persistent attacker will find it and exploit it until they’ve achieved their objective.

The second intrusion used a coercion trick known as PetitPotam coupled with the ESC8 vulnerability. How does this relay attack transform a temporary NTLM login into a lasting, high-privileged identity?

The second case shows an even more aggressive technical maneuver. When the attackers realized the ESC1 template weakness wasn’t available, they switched to a coercion technique called PetitPotam. This essentially tricks a domain controller into trying to authenticate with a machine the attacker controls. The attacker then takes that authentication attempt—an NTLM login—and “relays” it to the AD CS web enrollment page. This is the ESC8 abuse. Since the web enrollment page didn’t require a password or have proper protections, it accepted the relayed login and issued a certificate for the domain controller itself. The attacker then used that certificate with Kerberos (PKINIT) to get a ticket-granting ticket. Suddenly, they weren’t just a guest on a workstation; they were operating with the identity of the domain controller itself. It’s a masterful transformation of a fleeting network handshake into a permanent, ironclad identity that provides total control over the domain. Fortunately, in this instance, the organization was monitoring for ESC8 alerts and isolated the workstation before the attacker could move further, proving that fast response can win the day.

Given that many of these techniques have been known since 2021, why are they still proving so successful in 2026, and how do newer versions of Windows Server change the defensive landscape?

It’s a classic case of the “long tail” of infrastructure. Even though researchers disclosed ESC1, ESC8, and PetitPotam back in 2021, many organizations are still running legacy configurations or have migrated those old, insecure settings into newer environments. Security is a journey, and AD CS is often “set and forget” infrastructure that people are afraid to touch for fear of breaking logins. However, we are seeing progress. Microsoft turned on Extended Protection for Authentication (EPA) by default in Windows Server 2025, which significantly mitigates the ESC8 relay attacks. But if you’re still on Server 2019 or 2022, you have to go in and turn those protections on manually, as outlined in KB5005413. The reason these attacks work in 2026 is that attackers know defenders are overwhelmed. They know that while we’re looking for the “new” 0-day, the old, unpatched configuration on a forgotten certificate template is sitting there like a ticking time bomb.

What is your forecast for the evolution of identity-based social engineering over the next few years?

I expect we will see a much tighter integration between deep-fake technology and these help-desk impersonation schemes. Imagine receiving a Teams call where the “IT guy” not only has a support-themed name but also the voice and face of a person you actually recognize from the company directory. We are moving toward a world where “visual and auditory trust” is no longer a valid metric for security. Organizations will have to move toward a “zero-trust” communication model, where no remote-access session is granted without a secondary, out-of-band verification—perhaps a physical hardware token or a pre-shared secret that changes daily. The battleground has moved from the firewall to the identity provider, and if we don’t start treating our certificate services with the same level of scrutiny as our most sensitive databases, we’re going to continue seeing these devastating domain-level compromises. High-fidelity logging and certificate auditing won’t just be “nice to have”; they will be the only way to catch an attacker who has already legally “become” your administrator.

Explore more

How Can AI and CRM Architecture Drive Small Business Growth?

Navigating the complexities of the modern commercial landscape requires more than just grit and a good product because the sheer volume of digital signals generated by every customer interaction can quickly drown an unprepared entrepreneur. In the current market, the difference between a thriving enterprise and a stagnant one often comes down to how effectively a company manages its invisible

Canadian HR Leaders Unlock Strategic Value in Payroll Analytics

Hidden deep within the digital spreadsheets of every major Canadian corporation lies a sleeping giant of data that could fundamentally reshape the future of the modern workforce if only leaders knew how to wake it up. While most organizations view the biweekly distribution of funds as a repetitive administrative burden, forward-thinking Human Resources executives are beginning to see these transactions

Modern Marketing Automation Tools for Small Business Growth

Success in the current high-velocity marketplace is no longer defined by how many hours a founder works, but rather by how effectively they can delegate the mundane to sophisticated digital systems. Small business owners frequently find themselves trapped in a repetitive cycle of manual administrative tasks that ultimately stifle creative and strategic growth. With the modern marketplace demanding instant responses

How Does Event-Driven Marketing Solve Two-Way Messaging?

When a customer types the word “STOP” into a mobile messaging thread only to receive a promotional discount code three seconds later, the resulting frustration erodes years of brand loyalty in a single heartbeat. This specific moment of failure highlights the fundamental gap between the promise of modern customer engagement and the reality of legacy infrastructure. In 2026, where instant

How Will AI Integration Reshape B2B Marketing Budgets by 2027?

The Strategic Shift Toward an AI-Driven Financial Landscape The traditional boundaries of B2B marketing are dissolving as organizations reallocate massive amounts of capital toward integrated artificial intelligence systems that promise unprecedented scale. While many early adopters once viewed these technologies as simple shortcuts to lower overhead, the reality in 2026 is that AI is driving a profound redistribution of resources

Fill out the form below